Norm

1 Norms in Mathematical Analysis

1.1 Basic definition and axioms

In a vector space \(X\) over \(\mathbb{R}\) or \(\mathbb{C}\), a norm is a function \(\|\cdot\|:X\to[0,\infty)\) such that for all \(x,y\in X\) and scalars \(\alpha\),
1. Nonnegativity and definiteness: \(\|x\|\ge 0\) and \(\|x\|=0\) iff \(x=0\).
2. Homogeneity: \(\|\alpha x\|=	\alpha	\,\|x\|\).
3. Triangle inequality: \(\|x+y\|\le \|x\|+\|y\|\).

These axioms encode the idea that norms measure “size” in a way compatible with scaling and addition.

1.2 Relationship between norms and metrics

Every norm induces a distance function (a metric) on \(X\) by defining \[

d(x,y)=\|x-y\|.

\]

The norm axioms translate directly into the metric axioms: symmetry follows from \(\|x-y\|=\|y-x\|\), nonnegativity is immediate, and the triangle inequality becomes the usual triangle property for \(d\). This link allows analytic notions such as convergence, continuity, and compactness to be studied through either algebraic or geometric data.

1.3 Examples of norms on standard spaces

1.3.1 Absolute value on \(\mathbb{R}\) and \(\mathbb{C}\)

On \(\mathbb{R}\) the absolute value \(	x	\) is the standard norm. On \(\mathbb{C}\), the modulus \(	z	\) serves the same role, providing a notion of distance in the complex plane.

1.3.2 \(p\)-norms on \(\mathbb{R}^n\) and \(\mathbb{C}^n\)

For \(x=(x_1,\dots,x_n)\) and \(1\le p<\infty\), the \(p\)-norm is \[

\|x\|_p=\left(\sum_{k=1}^n	x_k	^p\right)^{1/p}.

\] For \(p=\infty\), one uses \[

\|x\|_\infty=\max_{1\le k\le n}	x_k	.

\] These norms reflect different ways of aggregating coordinate magnitudes and are central in finite-dimensional analysis.

1.3.3 Supremum norm on bounded functions

If \(X\) is a space of bounded functions on a set \(S\), the supremum norm is \[

\|f\|_\infty=\sup_{s\in S}	f(s)	.

\] This measures the largest pointwise deviation of a function and is a natural choice for uniform convergence.

1.4 Norm-induced notions of distance and size

With \(d(x,y)=\|x-y\|\), a sequence \((x_n)\) converges to \(x\) exactly when \(\|x_n-x\|\to 0\). A set of radius \(r\) centered at \(x\) can be described as

\[

\{y\in X:\|y-x\|<r\},

\] so balls replace intervals or neighborhoods. In analysis, this provides a unifying framework: “distance” is derived from algebraic structure rather than introduced separately.

2 Properties and Constructions

2.1 Homogeneity and triangle inequality

2.1.1 Equivalent formulations of the triangle inequality

The triangle inequality can be rearranged in several useful ways. For instance, using the substitution \(y=x-(x-y)\), one obtains \[

\|x\|-\|y\|	\le \|x-y\|.

\]

This inequality implies that the norm is a 1-Lipschitz function with respect to the induced metric, a fact frequently used to bound errors. Another common rearrangement is subadditivity in the form \(\|x\|=\| (x-y)+y\|\le \|x-y\|+\|y\|\).

2.2 Topological consequences of a norm

2.2.1 Open and closed sets in normed spaces

The metric induced by a norm determines which sets are open or closed. A set \(U\subseteq X\) is open if every point \(x\in U\) has some \(r&gt;0\) such that the ball \(B(x,r)\subseteq U\). Closed sets are complements of open sets, or equivalently sets containing all their limit points. Many familiar topological properties in analysis arise directly from these ball definitions.

2.2.2 Continuity of norm-related maps

Maps built from the norm, such as \(x\mapsto \|x\|\) or \(x\mapsto \|x-y\|\) for fixed \(y\), are continuous. More generally, if a function \(T:X\to Y\) between normed spaces satisfies \(\|T(x_n)-T(x)\|_Y\to 0\) whenever \(\|x_n-x\|_X\to 0\), it is continuous in the metric sense, and norm estimates often provide direct proofs.

2.3 Completeness and Banach spaces

2.3.1 Cauchy sequences and convergence

A sequence \((x_n)\) is Cauchy in a normed space if for every \(\varepsilon>0\) there exists \(N\) such that \(m,n\ge N\) implies \(\|x_n-x_m\|<\varepsilon\). Completeness concerns whether every Cauchy sequence converges to an element of the space.

2.3.2 Completion of normed spaces

A normed space \(X\) may be incomplete. Its completion is a complete normed space \(\overline{X}\) containing \(X\) densely, with a norm extending that of \(X\). Informally, completion fills “missing limit points” of Cauchy sequences, enabling analysis that relies on convergence within the ambient space.

2.4 Norm equivalence and comparability

2.4.1 Equivalent norms on finite-dimensional spaces

In finite-dimensional vector spaces, any two norms are equivalent: there exist constants \(c,C>0\) such that for all \(x\), \[

c\|x\|_1 \le \|x\|_2 \le C\|x\|_1.

\] Consequently, they produce the same notion of convergence and define the same open sets.

2.4.2 Impact of changing norms on convergence

In infinite-dimensional settings, changing norms can alter convergence behavior and continuity properties. A sequence that converges in one norm might fail to converge in another, unless the norms are equivalent or otherwise comparable through inequalities.

3 Normed Linear Spaces

3.1 Normed vector spaces (basic structure)

A normed linear space is a vector space equipped with a norm. Such spaces provide the setting for fundamental analytic concepts: continuity, differentiability in normed contexts, series convergence, and functional approximations.

3.2 Subspaces and restriction of norms

If \(Y\subseteq X\) is a linear subspace, restricting a norm from \(X\) to \(Y\) gives a norm on \(Y\). The induced metric on \(Y\) agrees with the one inherited from \(X\), so convergence and boundedness within \(Y\) can be studied using the same distance formula.

3.3 Quotient spaces and induced norms

Given a subspace \(M\subseteq X\), the quotient space \(X/M\) identifies vectors differing by elements of \(M\). An induced norm is defined by \[

\|x+M\|=\inf_{m\in M}\|x-m\|.

\] This measures how close \(x\) is to the equivalence class in a best-approximation sense and supports analysis of linear maps that “factor through” quotients.

3.4 Direct sums and product norms

3.4.1 Norms on \(X\times Y\)

For normed spaces \(X\) and \(Y\), one can define norms on the product space \(X\times Y\), for instance via \[

\|(x,y)\|=\|x\|_X+\|y\|_Y \quad\text{or}\quad \|(x,y)\|=\max(\|x\|_X,\|y\|_Y).

\] Each choice yields a valid norm and corresponds to a specific geometric perspective on pairs \((x,y)\).

3.4.2 Norms on \(X\oplus Y\)

When a direct sum is formed, vectors decompose uniquely as \(x+y\) with \(x\in X\), \(y\in Y\). Norms on the sum are typically constructed from norms on components, mirroring the product construction and preserving linear-algebraic operations alongside analytic estimates.

4 Special Types of Norms

4.1 Inner-product norms vs general norms

4.1.1 When a norm arises from an inner product

If a space has an inner product \(\langle\cdot,\cdot\rangle\), the induced norm is \[

\|x\|=\sqrt{\langle x,x\rangle}.

\] This produces a norm that is strongly tied to geometry: angles, orthogonality, and projections become meaningful and yield powerful inequalities.

4.1.2 Parallelogram law

A norm comes from an inner product precisely when it satisfies the parallelogram identity: \[

\|x+y\|^2+\|x-y\|^2=2\|x\|^2+2\|y\|^2.

\] This characterization is central for distinguishing genuinely inner-product-based geometry from more general normed geometry.

4.2 Operator norms in analysis

4.2.1 Norm of a linear operator

For a bounded linear operator \(T:X\to Y\) between normed spaces, its operator norm is defined as \[

\|T\|=\sup_{\|x\|_X=1}\|T(x)\|_Y.

\] This quantifies the maximal amplification \(T\) can impose and provides a standard measure of “size” for linear transformations.

4.2.2 Induced norms from vector norms

Operator norms depend on the norms chosen on \(X\) and \(Y\). Changing those norms generally changes \(\|T\|\), but comparisons between equivalent norms can preserve boundedness and continuity statements.

4.3 Seminorms and their relationship to norms

A seminorm satisfies nonnegativity, homogeneity, and the triangle inequality, but it may assign \(0\) to nonzero vectors. Seminorms occur naturally when identifying directions of “no measurement” or when measuring only certain components. Passing to a quotient by the kernel of the seminorm can convert the seminorm into a genuine norm.

4.4 Weighted and mixed norms

4.4.1 Weighted \(\ell^p\) norms

Weighted versions of \(\ell^p\) norms incorporate positive weights \(w_k\), for example \[

\|x\|_{p,w}=\left(\sum_{k} w_k	x_k	^p\right)^{1/p}.

\] They can emphasize or de-emphasize coordinates and are useful when different components represent different scales or importance.

4.4.2 Mixed norms in function spaces

In many applications, function spaces carry norms that combine integrals and norms in a structured way (e.g., integrating over one variable and taking an \(L^p\)-type norm over another). Mixed norms capture anisotropic behavior and appear frequently in estimates where different variables play different roles.

5 Norms in Functional Spaces

5.1 Sequence spaces (\(\ell^p\))

5.1.1 Convergence criteria in \(\ell^p\)

In \(\ell^p\) for \(1\le p<\infty\), convergence of sequences \(x^{(n)}\to x\) is characterized by \[

\|x^{(n)}-x\|_p^p=\sum_k	x_k^{(n)}-x_k	^p \to 0.

\]

For \(p=\infty\), convergence requires uniform control of coordinate differences: \(\sup_k	x_k^{(n)}-x_k	\to 0\).

5.1.2 Completeness of \(\ell^p\) spaces

\(\ell^p\) spaces are complete for \(1\le p\le \infty\). Thus they are Banach spaces, enabling robust limiting arguments for series expansions and approximation.

5.2 Function spaces (\(L^p\) and related norms)

5.2.1 Norms on \(L^p\) spaces

For a measure space, \(L^p\) consists of (equivalence classes of) measurable functions whose \(p\)-th power is integrable. The norm for \(1\le p<\infty\) is \[

\|f\|_p=\left(\int	f	^p\,d\mu\right)^{1/p},

\] while for \(p=\infty\) one uses the essential supremum norm. These choices reflect average behavior (\(p<\infty\)) versus uniform bound (\(p=\infty\)).

5.2.2 Almost-everywhere equality and well-definedness

In \(L^p\) spaces, two functions that differ only on a set of measure zero are identified as the same element. This ensures the norm is well-defined: changing a representative on a null set does not affect the integral defining \(\|f\|_p\), nor the essential supremum.

5.3 Sobolev-type norms (high-level overview)

5.3.1 Motivation from derivatives in \(L^p\)

Sobolev norms extend the idea of measuring a function not only by its size but also by the size of its derivatives in an \(L^p\) sense. The goal is to control regularity and enable compactness and convergence results for problems involving differential operators.

5.3.2 Typical uses in estimates

Sobolev-type norms are used to quantify how solutions behave and to bound quantities appearing in partial differential equation estimates. By encoding both magnitude and derivative information, they support inequalities that relate function values to gradients.

6 Applications in Analysis

6.1 Convergence, Cauchy criteria, and completeness

Norms provide the language for convergence tests: showing a sequence is Cauchy often suffices in complete spaces. Completeness turns abstract limiting procedures into concrete results, ensuring that approximations converge within the chosen function or sequence space.

6.2 Boundedness and Lipschitz-type estimates

Many estimates in analysis use inequalities of the form \[

\|f(x)-f(y)\|\le L\|x-y\|.

\] Such bounds imply Lipschitz continuity and control error propagation. Norms also provide the standard definition of boundedness for operators and functions, particularly when combined with supremum or integral norms.

6.3 Continuity and bounded linear maps

For linear maps between normed spaces, continuity is equivalent to boundedness. Specifically, a linear operator \(T\) is continuous exactly when there exists \(C\) such that \(\|T(x)\|_Y\le C\|x\|_X\) for all \(x\). The least such \(C\) is the operator norm.

6.4 Fixed-point methods and contraction mappings

6.4.1 Contractive conditions expressed via norms

The contraction mapping principle states that in a complete metric space, a self-map that strictly contracts distances has a unique fixed point. In normed spaces, contraction is expressed as \[

\|F(x)-F(y)\|\le q\|x-y\|

\] with \(0\le q<1\). The norm then directly provides the rate at which iterates converge to the fixed point.

7 Geometry of Normed Spaces

7.1 Unit balls and unit spheres

The unit ball \(\{x:\|x\|<1\}\) and unit sphere \(\{x:\|x\|=1\}\) describe the geometry induced by the norm. Different norms produce different shapes: for instance, \(\|\cdot\|_1\) yields a diamond-like ball in \(\mathbb{R}^2\), while \(\|\cdot\|_2\) gives a circle. These shapes influence approximation behavior and the structure of optimizers.

7.2 Convexity properties

7.2.1 Strict convexity and smoothness (conceptual)

A norm is strictly convex if the unit sphere contains no nontrivial line segments, which corresponds to uniqueness of best approximations in many settings. Smoothness refers informally to differentiability of the norm away from the origin, linking to how supporting hyperplanes touch the unit sphere. Together, these geometric traits affect stability and sensitivity of optimization problems.

7.3 Projection behavior (intuitive overview)

7.3.1 Best approximation ideas

In general normed spaces, projections can fail to behave like orthogonal projections in Hilbert spaces. However, one can still seek best approximations: given a point and a subspace, minimize \(\|x-y\|\) over \(y\) in the subspace. Existence and uniqueness depend on properties such as convexity and completeness, and the unit ball geometry often governs the outcome.

8 Further Directions

8.1 Normed vs metric spaces

While every norm yields a metric, not every metric arises from a norm on a vector space. Normed spaces emphasize algebraic structure—scaling and translation invariance—whereas general metric spaces can lack linearity. This distinction matters when transferring results: some theorems rely on the vector space operations, while others only need the metric.

8.2 Dual norms and the dual space

8.2.1 Defining the dual norm

For a normed space \(X\), the dual space \(X^*\) consists of continuous linear functionals \(f:X\to\mathbb{K}\). The dual norm is \[

\|f\|_{X^*}=\sup_{\|x\|_X=1}	f(x)	.

\] This definition measures the maximum action of \(f\) on unit vectors and is fundamental to functional analysis.

8.2.2 Hölder-type inequalities via duality

Duality often yields inequalities that bound pairings between elements of a space and its dual. The resulting statements generalize classical Hölder inequalities in \(L^p\) spaces and support estimate-driven proofs by translating norms into supremum formulas.

8.3 Extensions to seminormed and topological vector spaces

When only a seminorm is available, the induced structure is weaker: distinct vectors may have zero “distance.” Many constructions still work after passing to quotients by the seminorm kernel. More broadly, topological vector spaces generalize normed spaces by allowing convergence and continuity to be described via families of seminorms or more general neighborhoods, extending the analytic framework beyond metric settings.